The Research Least Squares Based On Ar Model Of Glucose Prediction

Recently, With the rapid development of our country’s economy and people’s material life level, Diabetes has become one of the health risks of chronic diseases, and there is an upward trend in the incidence of diabete. At present, the world has 400 million adults with diabetes in the number of diabetes patients. In China, More than 80% of China’s victim died of chronic diseases, Which represented by diabetes is far higher than the global average.Therefore, which is the great significance among the Strengthening the prevention and control of diabetes, Making the information intelligence technology to raise the level of diagnosis and treatment of diabetes, and warning of hypoglycemia to improve national health level.A good blood sugar prediction algorithm not only can be used for the daily blood glucose control, but also it can reducing the occurrence of hypoglycemia or hyperglycemia events, It is More important to slow diabetes complications; At the same time which can be used in conjunction with insulin pump, The control of insulin injection volume and injection time, more reasonable for real-time management of the human body blood sugar levels and the blood sugar control in a safe range.In blood sugar levels predicted mainly divided into two main direction in the field, one is the physical model based prediction, another is based on the data model prediction.Because the more complexly of human physiological mechanism and influence factors, It is more difficult to control and predict blood sugar levels. And the blood sugar prediction model based on data and the historical blood sugar levels, the real-time operation is more convenient through the appropriate data model to forecas, which has a larger increase prediction precision and universality.This thesis is based on the predictions of a data model, according to the features of historical blood sugar data, This thesis proposes a new prediction algorithm which is based on autoregressive AR model, with the AIC criterion to set the model order, and using the least squares method to solve the problem of time series prediction.It can use the dynamic blood sugar monitoring system(CGMS) for 72 hours of blood sugar monitoring, which record the change of the blood glucose fluctuations, Basing on the changes of blood sugar control diet, it is early opening or closing the pump operation, and it can change of insulin injection quantity and control blood sugar in a safe range.Finally, a numerical forecasting of 50 cases of patients with diabetes blood sugar, and Clarke analysis to evaluate the feasibility of the prediction error of grid, through AR prediction and monitoring of the comparison of real value, it show the algorithm can make blood sugar earlier forecast, and verified its effectiveness and suitability...